Color Consistency Recovery And Shadow Detection Of Grape Leaves Image Under Natural Illumination

With the development of modern agriculture, the application of digital image technology in the field of agriculture is becoming more extensive. In the field of agriculture, digital image processing technology is usually used to monitor the growth status of crops, derive information on plant diseases and insect pests in real time. Due to the incident angle of the sunlight, the plant leaves changes between mutual occlusion which often makes the image of plant leaf distort or shade producing some interference, such as color extraction, shadow region of lesion detection and segmentation. Therefore, this research is very significant because it’s aimed at restoring the image of plant leaf color, or rather weaken or eliminate the interference of the shadow and other factors on the leaf feature extraction. This research is further divided into two broad aspects:(1) Colors of the distinct images from the same leaf tend to look dramatically inconsistent, if under various illuminations or weather conditions,ten typical color constancy algorithms are used to process the grape leaf images and their effectiveness of color restoration is assessed in both methods: color difference and the angle error of the estimation source.The experimental results show that all algorithms work to some extent. Among them, 2nd-order Grey-Edge and Edge-based Gamut Mapping achieve the best results regardless of various illuminations, weather conditions or grape breeds. And, from the visual angle, the leaf colors of images processed by both algorithms are most similar to the ground truths. The results also indicate that: the 2kinds of color consistency algorithms are of great help to solve the problems of color variety in the images incurred by different imaging conditions, e.g.,and it can also be a good way to realize the color recovery of grape leaf images.(2) As regards the different materials and material reflectance in the visible and infrared bands, the main object of study was visible and NIR image of shadow detection processing of grape leaves in the visible light. At first, the visible and NIR images were normalized to produce the candidate shadow graph, the ratio of visible and NIR images were calculated and the shadow picture was obtained by multiplyingthe shadow candidate graph with the ratio of image; finally the binary shadow mask was gotten by adaptive thresholding.In order to carry on with the qualitative and quantitative evaluation, analysis and comparison of the algorithm proposed in this paper and two other advanced methods from the calculation accuracy ACC, Matthews correlation coefficient(MCC) and calculation of the running time. The test results showed that the method in this paper has high accuracy for all images; the mean value of ACC and MCC were higher than the other two methods while their standard deviations were lower than the other two methods. Because the shadow detection method is simple and the speed is the fastest among the same kind of advanced algorithm, it is therefore more accurate and reliable than the existing two algorithms.